July 7, 1999 CREATING AND ENFORCING NORMS, WITH SPECIAL REFERENCE TO SANCTIONS Richard A. Posner and Eric B. Rasmusen* Abstract Two central puzzles about social norms are how they are enforced and how they are created or modified. The sanctions for the violation of a norm can be categorized as automatic, guilt, shame, informational, bilateral-costly, and multilateral-costly. The choice of sanction is related to problems in creating and modifying norms. We use our analysis of the creation, modification, and enforcement of norms to analyze the scope of feasible government action either to promote desirable norms or to repress undesirable ones. We conclude that the difficulty of predicting the effect of such action limits its feasible scope. I. INTRODUCTION A norm is a social rule that does not depend on government for either promulgation or enforcement. Examples range from table manners and the rules of grammar to country club regulations and standard business practice. Norms may be independent of laws, as in the examples just given, or may overlap them; there are norms against stealing and lying, but also laws against these behaviors. The two kinds of rule reinforce each other through differences in the mode of creation, the definition of the offense, the procedure for administering punishment, and the punishments themselves. Laws are promulgated by public institutions, such as legislatures, regulatory agencies, and courts, after well-defined deliberative procedures, and are enforced by the police power of the state, which ultimately means by threat of violence. Norms are not necessarily promulgated at all. If they are, it is not by the state. Often a norm will result from (and crystallize) the gradual emergence of a consensus. Norms are enforced by internalized values, by refusals to interact with the offender, by disapproval of his actions, and sometimes by private violence. Norms are an attractive method of social control because a rule may be desirable but too costly a project for the state to undertake relative to the benefits. A rule against poor table manners, for example, is hardly suitable for embodiment in law. At the same time, norms have a number of drawbacks relative to laws. A norm is even more of a public good than a law, since no one person or political party can claim credit for creating a norm. And the cost of in?icting penalties for violating a norm cannot be financed by mandatory taxation and so must be shouldered voluntarily by those who enforce the norm. Because of these features of norms, it may seem obvious that norms would be undercreated and underenforced from a social standpoint. Yet we shall see that the underprovision thesis requires qualification. Norms, like laws, can be bad, so that the obstacles to their creation and enforcement may actually promote the social welfare. A related but subtler point is that because norms, once created, are difficult to uncreate, the stock of norms may be large even though the ?ow is small. The focus of the literature on norms has been on their importance and on their efficiency or inefficiency.1 Another theme has been their neglect by economists.2 We focus instead on how norms operate and so on the variety of sanctions that enforce norms, the degree of underenforcement associated with each type of sanction, and the difficulty of creating norms enforced by each type of sanction. The sanctions are various: 1. Automatic sanctions. The violator's action carries its own penalty because of its not being coordinated with the actions of others. Someone who drives on the wrong side of the road crashes into another car. 2. Guilt. The violator feels bad about his violation as a result of his education and upbringing, quite apart from external consequences. Probably most people in our society, though certainly not all, would feel at least somewhat guilty about stealing even if they believed they were certain not to be caught. 3. Shame. The violator feels that his action has lowered himself either in his own eyes or in the eyes of other people. In its most common form, shame arises when other people find out about the violation and think badly of the violator. The violator may also feel ashamed, however, even if others do not discover the violation. He can imagine what they would think if they did discover it, a moral sentiment which can operate even if he knows they never will discover it. Also, he may feel lowered in his own eyes, a "multiple self" situation in which the individual is both the actor and the observer of his actions.3 Because the focus is on the violator's status rather than on the violation, there can be shame even if there is no element of wrongdoing, no breach of a moral code; so one can be ashamed of doing something stupid that harms nobody, because behaving stupidly is failing to live up to one's self-image. This type of shame shades off into simple embarrassment. One can also be shamed (though the better word here would be "humiliated") for conduct that violates a moral code not one's own and so there is no question of guilt. During the Cultural Revolution in China, people paraded through the streets in dunce caps felt humiliated even if they disapproved of the regime and therefore felt no guilt at violating its norms. For simplicity, we shall treat humiliation as a form of shame, and shame itself as (1) a purely external sanction for (2) violations of the moral code. It is important to note, however, that even when viewed purely as an external sanction, that is, as the product of the actions or reactions of other people, shame (like guilt) is felt even if other people take no action. If a professor's arrest for patronizing a prostitute is publicized, he feels ashamed before his colleagues even though none of them mentions the arrest to him, perhaps because none of them mentions it-their silence, which is a reaction to the shameful act, is evidence that the act was indeed shameful and cannot just be laughed off. Neither guilt nor shame fits comfortably into the implicit psychological assumptions of the rational-choice model of human behavior. We treat them as brute facts, like food preferences, which constrain rather than explain economic theories of behavior. 4. Informational sanctions. The violator's action conveys information about himself that he would rather others not know. A student wears casual clothing to a job interview, unintentionally signaling that he doesn't really care about getting the job. 5. Bilateral costly sanctions. The violator is punished by the actions of and at the expense of just one other person, whose identity is specified by the norm. The expense to that person could be the effort needed to cause the violator disutility, or the utility that the person imposing the punishment loses by punishing him. Examples of what we are calling bilateral costly sanctions are where an adulterer is shot by a jealous husband and where the husband divorces his wife after discovering her adultery. 6. Multilateral costly sanctions. The violator is punished by the actions and at the expense of many other people. A divorced man finds that he is no longer invited to dinner in the community.4 A norm can be enforced by more than one sanction-indeed, by all six. A drunk driver weaves along the road and crashes into a bus, killing a child. He has wrecked his car, he feels guilty, he knows that all his neighbors look down on him, his employer discovers that he is an alcoholic, the child's parents condemn him, and he is ostracized by the entire community. All six sanctions have analogies in rewards for adherence to norms or for actions beyond the call of duty. A reward can take the form of a feeling of a duty well done, material gratitude from one or more people, the automatic gain from coordinated interaction with someone else, the signaling of desirable qualities, or the good opinion of others. We shall consider the sanctions one by one, but it may be useful at the outset to mention some general properties of optimal nonlegal sanctions. A sanction should be of the appropriate magnitude; if it is too severe, people will be overdeterred and refrain from efficient actions. A sanction should not be too costly; the cost will depend largely on how often it has to be administered, that is, on the frequency of violation of the norm. Because of the informality of enforcement, it is important that the application of a sanction not require too much information. If punishment will not be effective unless many people learn about the violation, there may be no punishment at all. II. SANCTIONS FOR VIOLATING NORMS Automatic sanctions. The classic example of an automatically enforced, or self-enforcing, norm is the convention of driving on the right side of the road. By violating the norm one increases the chances of an accident to oneself as well as to other people. Norms often are self-enforcing in the sense that the violator is punished automatically without the deliberate intervention of anyone. But the punishment may not be adequate-even in the case of driving on the right-because the violator considers the cost only to himself and not to other people. Drunk drivers sometimes drive on the wrong side of the road, even though they know they may hurt themselves, in part because they neglect the cost to other people. Coordination norms may also be suboptimal rules that are difficult to change, such as the way "though" is spelled in English. This, however, is a problem in altering rather than enforcing norms, so we defer it for now. Guilt. Guilt is an internal sanction, and so does not depend on the dissemination of information. It might be thought a type of automatic sanction, since the violator perceives it as a cost to himself, like the risk of being injured if he drives on the wrong side of the road. It differs from a simple automatic sanction, however, in requiring investment to create it. People must choose a level of effort to instill the potential for feeling guilt into other people (sometimes even themselves), as well as the amount of guilt to instill. This illustrates our earlier point that economics can "use" the concept of guilt even if it cannot explain it. The sense of guilt may be innate, but it is also developed by formal schooling, purposive moral in?uence by parents and relatives, and, possibly most important, the examples offered by both adults and peers. Parents have an interest, both selfish and altruistic, in instilling a certain amount of guilt in their children. (So they may instill too much guilt, from the standpoint of the child's long-run self-interest, for offenses such as rudeness, or ingratitude, to one's elders.) A child who has a sense of guilt is more likely to conform to norms, and other people's knowledge of this will help him later in life by making him a more reliable transacting partner. Rational parents will choose, and employ the level of effort required to achieve, the degree of strength of the guilt sanction that will allow for efficient breach of the norm, since often it is advantageous to violate a norm, especially if others are adhering to it. Rational parents will also endeavor to maintain marginal deterrence and so refrain from trying to make a child feel as guilty about not brushing his teeth as about shoplifting. That a sense of guilt should be to the child's long-run advantage illustrates one of the central insights of game theory-that reducing the potential payoffs that a player obtains in certain contingencies can increase his equilibrium payoff by inducing other people to alter their behavior. This point is illustrated in Figure 1, which models the parental decision as to whether to instill guilt about wrongdoing in their daughter. When the daughter grows up, an employer will decide whether or not to hire her, and if she is hired she will decide whether or not to steal from the employer. Assume that the dollar value of the job to the daughter is 40, the benefit to the employer from the daughter's work is 50, the dollar value of both the gain to the daughter and the loss to the employer from theft is 30, and the theft cannot be detected. The daughter's net pecuniary gain from being hired and committing theft is then 70 (40+30), and the employer's payoff from hiring the daughter is -20 (50-40-30). The daughter suffers guilt pangs that subtract G from her utility if she steals-provided her parents had instilled guilt in her. The parents and daughter are further assumed to have the same interests and the employer can observe (perhaps from the daughter's school record, references, and personal deportment) whether the parents have instilled a sense of guilt in her. The equilibrium depends on the size of the guilt cost, G. If G<30, then the parents have no incentive to instill guilt, since whether they do so or not their daughter would steal if hired because the payoff to her from stealing, 70-G, would exceed 40. Knowing this, the employer will not hire her. If G>30, however, the parents would instill guilt in their daughter; she would refrain from stealing if hired; and so the employer would hire her because her payoff from honest working, 40, would exceed her gain from dishonest working and the employer would derive a net benefit from hiring of 10 (50-40). The daughter thus benefits from her parents' creation of guilt feelings (conscience) in her, assuming that the job with this employer is her best job opportunity. She obtains the job by having the potential to feel guilty, but she does not suffer any actual pangs of guilt because she's honest. By having reduced the payoff to their daughter from stealing on the job, the parents have benefited her because of the reaction of the third party, the employer, to their action. FIGURE 1: How Parents Can Make Their Children Better Off by Reducing the Children's Payoff from Illegal Behavior Notes: 1.Numbers indicate payoffs to parents-daughter (the parents and daughter are assumed to have the same payoff) and to employer, respectively. 2. Arrows denote equilibrium actions if G>30. Shame. When viewed as a purely external sanction, shame requires information but is not an informational sanction (our fourth category), because it can arise even if the violation does not convey any new information about the character of the violator. Often, however, what we call shame is partly an informational sanction: the man caught drunk driving is revealing that he probably drinks heavily on a regular basis, and it is his consciousness of people's reaction to the revelation that engenders the feeling of shame in him. This can be true even of purely internal shame; someone titillated by his first sight of sexually deviant pornography may be ashamed to discover inclinations within himself that he never suspected he had. Shame is often a byproduct of bilateral or multilateral sanctions. When people criticize a norm violator they are trying to impose a multilateral sanction, but the disutility of the criticisms to the violator, and thus the efficacy of the sanction, may be due entirely to shame. If the violator shrugs off the criticisms as a product of ignorance, malice, or envy, and in addition anticipates no bad effects on him from the reception of the criticisms by other people, the criticisms will fail as sanctions for the criticized act. Shame, like guilt, is a product, in part at least, of education, both formal and informal. Parents may have a greater incentive to instill shame in their children than guilt. The optimal strategy from the child's selfish (which may be the parents' altruistic) viewpoint is to violate norms but not get caught, and shame, to the extent that it is an external sanction and thus depends on information, provides the right incentives for that. Informational sanctions. If the violation of a norm is positively correlated with the possession of undesirable characteristics, people will be likely to punish the violator.5 The undesirable personal characteristic could be a tendency to violate the particular norm-as when a man beats a woman and is punished by not being able to find any other woman willing to associate with him-or to violate norms generally. The sanction is efficient. Not only does it impose significant costs on the violator; its cost to the enforcer is actually negative, because he averts a loss by avoiding dealings with a person who by his violation has revealed himself to be an unreliable transactional partner. The personal (or commercial) characteristic revealed by the violation may be only distantly related to the violation yet convey valuable information, as when a person arrested for assault is believed to be less likely to work hard for whoever employs him. Consider the following game: 1. Nature chooses 90 percent of workers to be "steady," with productivity p=x, and 10 percent to be "wild," with productivity p=x-y. 2. A worker decides whether to marry or not. Marriage adds utility u=m for a steady worker and utility u=-z for a wild worker. 3. The employer, observing whether the worker is married but not whether he is wild, offers him a wage wm or wu in competition with other employers, depending on whether he is married or not. The employer has no intrinsic reason to care whether the worker is married or not. Wild workers are less productive, but whether they are married has no effect on their productivity. The only significance of marriage for the employer is its informational value as a signal of steadiness. 4. The payoffs received are pw for the employer if he succeeds in hiring the worker, wu for a married employed worker, w for an unmarried employed worker, u for a married unemployed worker, and 0 for an employer who does not hire and for an unmarried worker who is not hired. What will the equilibrium be? Unlike many signaling models,6 here there is only a single equilibrium. If z is large enough (namely, greater than y), the employer will pay wages of wu=x-y and wm=x, the steady worker will get married, and the wild worker will stay single.7 The employer in this example might be "unthinkingly" obeying a norm of paying married workers more. Businessmen, like private individuals, follow many norms without inquiring into their rationality. In the example, following the norm is efficient even if no businessman understands its origin or rationale. And notice that unmarried workers will not marry in order to fool the employer into paying them a higher wage, because by definition these are the workers for whom the wage premium is less than the disutility (stemming from their "wildness") of marriage. Notice also in this example that subsidizing marriage not only would not raise productivity, but would lower it, by depriving employers of useful information about the marginal product of their workers. The same is true, however, of taxing marriage; obviously the marriage signal would be destroyed if marriage were taxed out of existence. And likewise if government forbade employers to use an applicant's marital status in making a hiring decision. This is an example of the danger of government's fiddling with norms. Informational sanctions may seem potentially too severe. A trivial violation of a norm, if the violation signals the offender's probable unreliability as a friend or business acquaintance, may precipitate ostracism that will impose a cost on the violator which exceeds the social cost of his breach. But this need not be excessive punishment, since it corrects an asymmetry of information and by doing so confers a social benefit that is distinct from its deterrent effect. Bilateral costly sanctions. These require only a minimal dissemination of information; the designated punisher is the only person who needs to learn of the violation. The difficulty lies in getting the punisher to carry out the punishment, given that it is costly to him, unlike the previous cases that we have been considering. Providing motivation for the punishment may require second-order internal norms, or a system of punishing nonpunishers. Corsican norms of vengeance, for example, supplemented bilateral costly sanctions with multilateral ones by ostracising anyone who was unwilling to carry out his duty of avenging an injury.8 Punishment can be facilitated by reducing its costs, as in the preceding example, where the net cost of punishing, which is the difference between the cost of punishing and the cost of not punishment, is reduced by imposing a cost on the nonpunisher. The costs of punishment can also be reduced by releasing the punisher from the ordinary sanctions, formal or informal, for the punishing behavior and relying on his vengeful emotions, which may exist even without second-order norms, to create a perceived benefit of vengeance in excess of the now reduced costs.9 Ordinarily, a person who has insulted someone in public is sanctioned, but if his insults are punishment for the other person's norm violation, the insults are excused. Children early discover the not always credible excuse: "But he started it!" Multilateral costly sanctions. Multilateral punishments require more information than bilateral ones. The free-rider problem is exacerbated because more people are involved but ameliorated because each punisher's cost of punishment can be less for the same total deterrent. The Amish practice of shunning offenders against church rules is an example. Like other Mennonites, the Amish have norms both against violence and against going to court. Excommunication, with resulting ostracism, is therefore the maximum punishment.10 But it is effective because the Amish constitute an isolated subculture from which exit is costly. Ostracism lies at the end of a spectrum the other end of which is the "dirty look"-an expression of disapproval with no tangible penalty attached. It is remarkable how sensitive people are to manifestations of disapproval even by strangers and even when unaccompanied by any implicit or explicit threat. Of course, this form of the multilateral costly sanction works only when the benefits of violating a norm are slight; but it does work then, because the cost of this form of sanction to the sanctioner is low. III. THE CREATION AND DESTRUCTION OF NORMS A norm is a public good. Although exceptions can be imagined, ordinarily a norm is nonrivalrous, because its cost does not rise if more people use the norm; and it is nonexcludable, because people who do not contribute to its enforcement cannot be denied its benefits. The norm is therefore in danger of being underproduced. By the same token, once a norm is established, it is hard to change, because norm innovation is also a public good. Creating a norm requires promulgation of the norm and creation of sanctions for its violation. Eliminating a norm requires promulgation, too, and also the destruction of the expectations and tastes that support the sanctions for its violation-a process of taste-changing that may be as costly as their creation in the first place. Changing a norm, which requires elements of both destruction and creation, can be the most difficult trick of all. Consider changing a self-enforcing coordination norm, such as a rule of language. If "goodbye" became an inefficient form of farewell, individuals could shift to "take care" without cost. Any one person who changed would be able to convey the same meaning as before, and would not be thought worse of by other people. Many writers are now shifting from using "he" as the indefinite pronoun to "he or she," "he/she," or "she." The writer's meaning is still apparent, as in the case of shifting from "goodbye" to "take care," but in the "he/she" case the violation of the grammatical norm distracts the reader and slows down communication. The rapidity with which the norm has nevertheless changed is a result of the felt value to many people of signaling a belief in feminist goals. It has not changed all the way, however, because some writers wish to signal their rejection of what they take to be an extremist feminist ideology. In both these examples, altering an existing norm is feasible because it can be done gradually and hence without centralized direction. A coordination norm that cannot be changed gradually would be reversing traffic signals so that red meant "go" and green meant "stop," or changing from driving on the right to driving on the left. Such a norm change would have to be adopted by everyone at once, unlike minor language changes, in order to avoid enormous transition costs. The only- but considerable-cost of the norm changes in our traffic examples lies in the habitual character of compliance with these norms. Another familiar example is the ordering of keys on a typewriter or computer keyboard. The keyboard cannot gradually evolve, letter by letter, into a new arrangement. In the age of the computer one can remap the keys into different symbols one by one, so that the capital costs would be slight, but for someone who touchtypes the disruption from a change in even one key is severe because of habituation to the existing pattern. If the transitional costs to a new norm are high enough, we have the phenomenon of a norm "trap," meaning that society is stuck with a suboptimal norm because of the costs of changing it.11 Trying to change from one language to another, as opposed to changing usage within a single language, would present a similar problem on a vaster scale.12 The cost would be enormous and the benefits highly diffuse even though they might be enormous too. National languages do change, but ordinarily it is through an intermediate stage of bilingualism. What has been more common is reform in writing, as in the simplification of Chinese characters in the twentieth century,13 the replacement of the Gothic alphabet by the Roman in Germany after World War II,14 and the increasing substitution in Korean writing of the highly phonetic Korean "alphabetic syllabary" for Chinese characters.15 But in the Chinese and Korean cases, the change was effected by government, just as in the case of Sweden's overnight change from driving on the left to driving on the right.16 The problem of transitions (path dependence) is not special to norms, however; it arises whenever standardization is desirable, as in the typewriter keyboard, railroad gauges, the size of screws, and the color of traffic signals. But it explains why often it is desirable that norms should change slowly even when the change is in the direction of a clearly superior norm. Because the superior norm may take time to phase in, there may be a rent in the fabric of social control if the older, inferior norm disappeared before the transition was complete. Consider the gradual decline of vengeance as an extralegal normative system for deterring and punishing crime. The decline paralleled the increasing sophistication and efficacy of law's methods of crime prevention. Had the vengeance norm collapsed suddenly, anarchy would have resulted because the legal methods of crime control (involving police, judges, lawyers, and so on) were not yet highly developed. Dovetailing the fading and the emerging norm is difficult, however. Rapid change in the social environment may make norms dysfunctional before enough time has passed for the normative system to adapt fully. The overdeveloped sense of honor in white males of the Old South (a residue of a vengeance culture), the similar "macho" values of poor young black males in our cities today, and the surprisingly resilient "coolness" of cigarette smoking are all examples of norms that apparently are dysfunctional under current conditions but that nevertheless persist. Without public intervention, many norms change only gradually or not at all. (This is our earlier point that the destruction as well as the creation of norms is a public good, so that it is possible for the ?ow of norms to be too small but the stock of norms too large.) This is only a tendency; Cooter gives the counterexample of a relatively rapid change in tribal norms in New Guinea which formerly allowed the "big man" to alienate property, a norm that became dysfunctional once an active land market developed.17 At the opposite extreme, if frequent change is itself a norm, as in women's fashions, the impression of frequent and easy norm change is misleading. The surprise would be if the "change norm" itself changed suddenly-if one year women's fashions were identical to the previous year's. Where faddism is the norm, the rapid succession of fads indicates norm stability, not instability. Norms enforced by guilt and shame are particularly difficult to create or to change. (This is a factor in the resistance of honor cultures to norm change: these cultures rely heavily on shame to induce norm compliance.) Guilt and shame are heavily in?uenced by social conditioning, which is not quickly or easily altered either by individuals or by governments. Nongovernmental organizations may be more effective than either individuals or governments in this regard, but it is not clear whether a society that gives ample scope to norm-changing organizations will have more or less norm creation and stability. Religious pluralism, for example, has a complex effect on norms. It facilitates norm shopping, both because a person can find a religion that does not constrain his activities and because he can find one that constrains them in a way that he finds desirable. Belonging to a church that punishes deviance can be as useful to one's long-term welfare (in quite materialistic terms) as having parents who instill guilt. And a religious sect in a pluralist society can provide an alternative community to the society at large, as in the case of the Amish. But while norm competition, which religious freedom promotes, encourages the creation of norms, it also facilitates their destruction. Understanding how norms change aids in understanding how they are created. When incremental change is feasible, even complex norms can evolve from meager beginnings, given enough time. Comparing the scene depicted on Achilles' shield in Book XX of the Iliad with modern norms of judicial behavior, one gets a glimpse of how those norms could, over a period of more than 2,500 years, have evolved from rudimentary procedures. The shield depicts informal, voluntary arbitration before a lay tribunal. This is remote from modern litigation (though not so remote from modern arbitration), yet it is easy to trace a step by step evolution from the ancient to the modern practices of dispute resolution. Heavy sanctions for violating a norm affect both norm change and who becomes a norm innovator. Kuran shows how norms that appear to be very strong can suddenly evaporate, enforced as they are by people's beliefs that other people will continue to enforce them.18 And Tocqueville wrote that revolutionary crowds are led literally by madmen: people who through psychological illness are willing to risk danger and disgrace, or in other who are less responsive to sanctions.19 Some norms are easier to create than to enforce. It is relatively easy to create coordination norms, since their punishments are automatic, although, as we noted earlier, such punishments are unlikely to be optimally severe, because the violator will ignore the costs of the violation to everyone but himself. On the other hand, a norm that requires multilateral punishment is hard to create in the first place, but once created it may be cheap to enforce (because little is required of each punisher) and the problem of cost externalization by the violator will not arise. The diversity of norms along such dimensions as efficacy, durability, mode of enforcement, and conformity to overall social goals limits the scope of useful generalization. Such questions as whether the United States has too few or too many norms, or whether norms are a good thing or a bad thing, or whether they "work" in a society as heterogeneous and individualistic as that of the United States are either unanswerable or misconceived. (To see this, substitute "law" for "norms.") A more particularistic analysis is required. IV. NORMS, LAWS, AND GOVERNMENT Norms are particularly effective devices for social control, relative to law, when individual violations (though perhaps not aggregate violations) are too trivial, or the difficulty of proving guilt too great, to justify the expense of trials, police, and prisons. But the sanctions for violating norms are often too weak to deter all people from many offenses, while norm creation is too slow to provide for all the rules necessary for the governance of society-so laws have their place too. Government can provide supplemental punishments if the informal sanction for violating a norm is inadequate. It does this for theft, for example. Indeed, it is a common observation that norms are more important than laws in deterring theft, though one might question whether the norms would long survive elimination of the laws. Some norms would-commercial norms arose before they were enforced by courts. The government can often leave the promulgation of a desired norm to the private sector while providing sanctions for violations once the norm is established. Legal sanctions for norm violations are also important because many people are impervious to informal sanctions. They lack guilt and shame, do not mind ostracism (because they have no valuable transactional opportunities regardless of their norm compliance), or have no reputation to lose, but they are still vulnerable to the law's tangible sanctions. An additional role of government in relation to norm enforcement is provide information. If multilateral or shame sanctions are to work, the violation must be publicized. With any nonautomatic sanction except guilt, incorrect information is a problem. Innocent people may be punished even if they did not really violate the norm. This is not just a problem for the information sanctions. Mistaken creation of shame is also possible. A person does not like to be found in a compromising situation even if he knows himself to be fully innocent and thus does not experience a feeling of guilt. The legal process is designed to minimize the likelihood of erroneously imposing formal legal sanctions. Elaborate protections of the innocent are not necessary with regard to most of the conduct that norms rather than laws regulate, however, because the sanctions for violation are not severe. But when the norm violations are punished by especially severe extralegal sanctions, a public hearing to correct the erroneous imposition of the sanction is warranted. This is the economic rationale for the provision of legal remedies against defamation. Defamation is an information sanction, discouraging people from dealing with the defamed person. But if the penalty for slander (oral defamation, as opposed to written, which is governed by libel law) is too high, people will be afraid to gossip for fear that they do not have the story exactly right. Gossip is an important facilitator of sanctions for violating norms.20 A third point (the first, remember, was supplemental punishment and the second information) is that government should be careful about interfering with norm sanctions. Sometimes just staying out of the way is the best policy. If bilateral sanctions such as shotgun marriages are to work, the law has to relax its monopoly on force.21 Preventing ordinary citizens from carrying concealed weapons may increase violent crime by hindering the bilateral costly sanction of self-defense.22 Fourth, the government can supply incentives for administering private sanctions for violating a norm. Consider the enforcement of contracts that provide for arbitration of disputes arising under them. A contract is a set of norms constructed by two parties. If the law enforces arbitration awards (as it does), it gives legal backing to private norm formation and enforcement. Fifth, the government can foster the creation of norms. In the case of coordination norms, the government can promulgate the new norm, as in the examples we gave earlier. When guilt and shame are the sanctions, the government can help to instill these in children and adults alike. But because moral and intellectual education can work at cross-purposes, an increase in the resources devoted to education is not a dependable means of promoting governance by norms. Intellectual education weakens norms by revealing alternative norms and by providing the tools for the student to use in creating his or her own moral principles.23 "Liberal" education is "freeing" education and one of those freedoms is from norms. Indeed, if education is more effective at promoting individualistic thinking than at inculcating norms, the net effect of education may be to reduce the normative regulation of society. Whether education is effective at either task is, of course, the despair of teachers at all levels, so it may be that the net effect is small whichever effect preponderates. Nevertheless, what appears to be a long-term movement away from regulation by norms and toward regulation by law may re?ect an inverse relation between a society's level of education and the efficacy of regulation by norms. Moreover, education and income are positively correlated; privacy is a superior good; and privacy reduces the efficacy of norms by depriving neighbors, acquaintances, gossips, and scandal sheets of the information needed for shame, informational, and mutilateral sanctions.24 Sixth, the government has a role to play in combating bad norms. It can do this by diminishing the benefits of compliance with such norms by creating effective legal remedies for deliberate injuries, which reduces the benefit of a vengeance norm based on (for example) personal honor. Or it can increase the cost of complying with the norm, most simply by affixing a legal penalty such as making dueling a crime. But since the heart of an honor-based vengeance system is a readiness to act without regard to the balance of costs and benefits, increasing the costs of compliance with a bad norm is not so simple as it seems. Once a legal penalty is affixed, compliance with the norm may even more effectively signal the dueler's honor. However, honor implies indifference only to certain costs, in much the same way that indigence may make one indifferent to uncollectable fines, but not to imprisonment. The proper "currency" in which to punish dueling is to make duelling dishonorable, as by disqualifying the dueller from the public offices that a man of honor is duty-bound to fill.25 Dueling is too "safe" an example, however, of a bad norm, since no one in our society approves of dueling. One danger of governmental intervention to promote "good" norms and crush "bad" ones is that norms are frequently contested, so that government may find itself taking sides in an unedifying struggle between interest groups. The Constitution provides a partial barrier to such intervention through its guaranties of freedom of religious, speech, and association. Whether government has the capacity to carry out the functions sketched above may be questioned, in light of the fine tuning required. Most often it may be best for the government to "keep out of the way" and allow the private regime of norm creation and enforcement to operate unhindered. But obviously the government cannot be kept out of the norm business by any feasible constitutional rule; and we emphasize that intervention which takes the form of supplementing and regulating rather than supplanting norm sanctions is likely to be less intrusive than many other forms of government regulation. Norms provide a private, decentralized, and competitive alternative to governmental control of social behavior. * Posner: U.S. Court of Appeals for the Seventh Circuit, 219 South Dearborn Street, Chicago, Illinois 60604. Office: (312) 435-5806. Fax: (312) 435-7545. Email: richard_posner@ca7.uscourts.gov. Rasmusen: Indiana University, Kelley School of Business, Rm. 456, 1309 E 10th Street, Bloomington, Indiana, 47405-1701. Office: (812) 855-9219. Fax: (812) 855-3354. Email: Erasmuse@Indiana.edu. Web: Php.Indiana.edu/~erasmuse. We thank T. Lynn Fisher, Kelly Lear, and Brady Mickelsen for their helpful assistance, and Robert Ellickson, Richard McAdams, Michael McGinnis, J. Mark Ramseyer, two anonymous referees for the International Review of Law and Economics, and participants in Indiana University's Workshop in Political Theory and Policy Analysis for their helpful comments. 1 The literature is large and growing. One branch examines business dealings. The seminal article, sociological rather than economic in emphasis, is Stewart Macaulay, "Non-Contractual Relations in Business," 28 American Sociological Review 55 (1963), and illustrative recent contributions are Lisa Bernstein, "Opting Out of the Legal System: Extralegal Contractual Relations in the Diamond Industry," 21 Journal of Legal Studies 115 (1992), and Robert D. Cooter, "Decentralized Law for a Complex Economy: The Structural Approach to Adjudicating the New Law Merchant," 144 University of Pennsylvania Law Review 1643 (1996). Another branch, which develops an economic theory of social custom, begins with two articles by George Akerlof: "The Economics of Caste and of the Rat Race and Other Woeful Tales," 90 Quarterly Journal of Economics 599 (1976), and "A Theory of Social Custom, of Which Unemployment May Be One Consequence," 94 Quarterly Journal of Economics 749 (1980). A useful survey of theoretical and experimental work on the role of norms in solving collective-action problems is Elinor Ostrom, "A Behavioral Approach to the Rational-Choice Theory of Collective Action," 92 American Political Science Review 1 (1998). Much recent work focuses on the relation between private norms and public laws. See, for example, Robert Cooter, " Models of Morality in Law and Economics: Self-Control and Self-Improvement for the "Bad Man" of Holmes," 78 Boston University Law Review 903 (1998); Robert Ellickson, Order without Law: How Neighbors Settle Disputes (1991); Avery Katz, "Taking Private Ordering Seriously," 144 University Pennsylvania Law Review 1745 (1996); Eric A. Posner, "The Regulation of Groups: The Influence of Legal and Nonlegal Sanctions on Collective Action," 63 University of Chicago Law Review 133, 155-161 (1996); Eric A. Posner, "Law, Economics, and Inefficient Norms," 144 University of Pennsylvania Law Review 1697 (1996); J. Mark Ramseyer, "Learning to Love Japan: Social Norms and Market Incentives," 34 San Diego Law Review 263 (1994);"Symposium, Social Norms, Social Meaning, and the Economic Analysis of Law," 27 Journal of Legal Studies (1998). 2 This is an old story: Thomas Macaulay and James Fitzjames Stephen attacked James and John Stuart Mill, respectively, for overemphasizing laws relative to norms as explanations for differences in culture and behavior. Thomas Macaulay, "Mill on Government," Edinburgh Review (March 1829) and "Westminster Reviewer's Defense of Mill," Edinburgh Review (June 1829); James Stephen, Liberty, Equality, Fraternity , chapters 1 and 4 (1874). John Stuart Mill, however does emphasize the tyranny of public opinion, thus tacitly recognizing the importance of informal norms in shaping behavior. John Stuart Mill, On Liberty (1859). 3 On multiple-self analysis, see, for example, Richard A. Posner, "Are We One Self or Multiple Selves? Implications for Law and Public Policy" 3 Legal Theory 23 (1997). 4 Our categories of guilt and shame correspond to "first party" sanctions in Ellickson's terminology; our bilateral costly to his "second party"; and our multilateral costly to his "third party." Ellickson, note 1 above, at 130-131. He does not include what we are calling the automatic and the informational sanctions in his taxonomy because they are not methods of enforcing norms as he defines "norm." For other discussion of definitional issues (which we'll largely duck), see Richard McAdams, "The Origin, Development, and Regulation of Norms," 96 Michigan Law Review 338 (1997). 5 See, for example, Eric A. Posner, "Symbols, Signals, and Social Norms in Politics and the Law," 27 Journal of Legal Studies 765 (1998); Eric Rasmusen, "Stigma and Self-Fulfilling Expectations of Criminality," 39 Journal of Law and Economics 519 (1996). 6 See Eric Rasmusen, Games and Information, ch. 10 (2d ed. 1994), and next footnote. 7 The analysis is more complicated, however, if zz>.9y, wild workers will split in their behavior, in a mixed-strategy equilibrium. The employer will pay wages of wu=x-y and wm=x+y+z; the steady worker will get married; the fraction f=.9(y-z)/z of the wild workers will marry; and the employer will earn zero profits when he hires a married worker-i.e., f solves .9x/(.9+.1f)+.1f (x-y) /(.9+.1f)-(x-y+z)=0. A further complication, but not one that would alter the implications of our analysis, is that efficiency might require that the different types of worker (steady and wild) be sorted into different types of job (for example, management and sales). 8 Jon Elster, The Cement of Society 127 (1989). 9 See Richard A. Posner, The Economics of Justice, ch. 8 (1981); Jack Hirshleifer, "On the Emotions as Guarantors of Threats and Promises, in The Latest and the Best 307 (John Dupre ed. 1987); Robert H. Frank, Passions within Reason: The Strategic Role of the Emotions (1988). 10 For example, Andrew Yoder, having been ostracized by church orders, sued four church leaders for conspiracy to deprive him of civil liberties. He won, and upon the defendants' refusing to pay damages one of the church leader's farms was sold by the sheriff. The judgment might well have been reversed on appeal, but the norm against going to court seems to have prevented the hiring of defense counsel and the filing of an appeal. John Yoder, "Caesar and the Meidung," 23 Mennonite Quarterly Review 76 (1949). For a theoretical discussion of costly ostracism in a repeated game, see David Hirshleifer and Eric Rasmusen, "Cooperation in a Repeated Prisoner's Dilemma with Ostracism," 12 Journal of Economic Behavior and Organization 87 (1989). 11 The claim that the QWERTY keyboard is inefficient has, however, been greatly undermined by S. Liebowitz and Stephen Margolis, "The Fable of the Keys," 33 Journal of Law and Economics 1 (1990). See also Nicholas Economides' networks website at raven.stem.nyu.edu/networks. In any event, no one doubts that the transition costs of adopting a new keyboard would be considerable. 12 See Richard Adelstein, "Language Orders," 7 Constitutional Political Economy 221 (1996). 13 See, for example, Insup Taylor and M. Martin Taylor, Writing and Literacy in Chinese, Korean and Japanese, ch. 8 (1995); Language Reform in China: Documents and Commentary (Peter J. Seybolt and Gregory Kuei-ke Chiang eds. 1979). 14 Kenneth Katzner, The Languages of the World 70-71 (new ed. 1995). 15 See Taylor and Taylor, note 13 above, pt. 2. The Korean alphabet, introduced by the government early in the fifteenth century, shows that the reform of writing is not a recent invention. Failed attempts to change languages abound. Consider Ireland's attempt to expand the use of Gaelic, or Canada's attempt to spread the use of French. There have been notable successes, but most have been the result of conquest; an exception is the revival of the dead language Hebrew as the national language of Israel. 16 See "Sweden Tells Traffic to Keep to the Right," Business Week, Sept. 2, 1967, p. 26. 17 Cooter, note 1 above. 18 Timur Kuran, Private Truths, Public Lies: The Social Consequences of Preference Falsification (1995). 19 Alexis de Tocqueville, Souvenirs 195 (1978 [1850]): "I have always thought that in the revolutions and especially in democratic revolutions, madmen, not those to whom one gives the name metaphorically, but real ones, have played a very considerable political role. This at least is certain, at least, that a half-crazy false messiah in this time soften goes on to success." 20 See Sally Merry, "Rethinking Gossip and Scandal," in Toward a General Theory of Social Control, vol. 1, p. 271 (Donald Black ed. 1984). 21 On the importance of this norm, see George Akerlof, Janet Yellen, and Michael Katz, "An Analysis of Out-of-Wedlock Childbearing in the United States," 111 Quarterly Journal of Economics 277 (1996). They find that the change in the 1960s in the norms governing births out of wedlock was due to the rise of birth control and abortion, which reduced the benefits of shotgun marriages. Most such marriages, of course, were not literally enforced by the threat of violence; nor do we suggest that permitting such threats would be warranted by the costs of out-of- wedlock births, substantial as those costs are. 22 John Lott and David Mustard, "Crime, Deterrence, and the Right to Carry Concealed Handguns," 26 Journal of Legal Studies 1 (1997). 23 As emphasized in Richard A. Posner, "Social Norms and the Law: An Economic Approach," 87 American Economic Review Papers and Proceedings 365 (May 1997); see also Richard A. Posner, The Problematics of Moral and Legal Theory 70-75 (1999). 24 See Yuval Tal, Privacy and Social Norms: Social Control by Reputational Costs (unpublished diss., University of Chicago Law School, 1997). 25 Lawrence Lessig, "The Regulation of Social Meaning," 62 University of Chicago Law Review 943, 971-972 (1995). Creating and Enforcing Norms Creating and Enforcing Norms Revised July 7, 1999